Fragmented ecosystems of the desiccated Aral Sea seek answers to the profound local hydrologically- and water-related problems. Particularly, in the Small Aral Sea Basin (SASB), these problems are associated with low precipitation, increased temperature, land use and evapotranspiration (ET) changes. Here, the utility of high-resolution satellite dataset is employed to model the growing season dynamic of near-surface fluxes controlled by the advective effects of desert and oasis ecosystems in the SASB. This study adapted and applied the sensible heat flux calibration mechanism of Surface Energy Balance Algorithm for Land (SEBAL) to 16 clear-sky Landsat 7 ETM+ dataset, following a guided automatic pixels search from surface temperature Ts and Normalized Difference Vegetation Index NDVI ( ). Results were comprehensively validated with flux components and actual ET (ETa) outputs of Eddy Covariance (EC) and Meteorological Station (KZL) observations located in the desert and oasis, respectively. Compared with the original SEBAL, a noteworthy enhancement of flux estimations was achieved as follows: – desert ecosystem ETa R2 = 0.94; oasis ecosystem ETa R2 = 0.98 (P < 0.05). The improvement uncovered the exact land use contributions to ETa variability, with average estimates ranging from 1.24 mm to 6.98 mm . Additionally, instantaneous ET to NDVI (ETins-NDVI) ratio indicated that desert and oasis consumptive water use vary significantly with time of the season. This study indicates the possibility of continuous daily ET monitoring with considerable implications for improving water resources decision support over complex data-scarce drylands.
12 January 2026 in IEEE Transactions on Geoscience and Remote Sensing
Autonomous feature extraction from high-resolution (HR) Remote Sensing (RS) imagery presents complex spatial–spectral heterogeneity and diverse land-cover patterns, challenging conventional single-architecture models that lack multiscale feature diversity and cross-domain generalization. Convolutional neural networks (CNNs) effectively capture localized hierarchical structures, whereas transformers model long-range contextual dependencies via self-attention. However, using these architectures independently restricts their complementary potential. This study proposes FWDNNet (Fused Weights Deep Neural Tokenization Networks), a fusion framework introducing a CNN-to-token conversion mechanism that transforms convolutional feature maps from heterogeneous encoders into transformer-compatible token sequences. Beyond standard concatenation, FWDNNet integrates a TensorDot-based fusion module to learn high-order cross-architectural interactions, while a shared decoding pathway maintains multiscale feature hierarchies and computational efficiency. Experiments across three diverse datasets show that FWDNNet achieves 58.2 ms inference latency with minimal cross-domain performance degradation (2.3–4.2% mIoU drop). Compared to conventional fusion strategies, FWDNNet attains 21% faster inference and 1.7–2.7% higher mIoU over baseline methods. See Materials at https://github.com/BoazGithub/FWDNNet for reproducibility.